TRANSPARENT AND ACCURATE OBESITY RISK CLASSIFICATION USING ENSEMBLE LEARNING AND LIME EXPLANATIONS

Abstract

The purpose of this method is to make sure that forecasts are clear and easy to use for planning healthcare, while also accurately assessing the risk of obesity. It makes use of the advantages of several classifiers, such as Random Forest, Gradient Boosting, and Logistic Regression, by employing a hybrid ensemble learning methodology. In order to address the complex nature of obesity, the methodology takes into account a wide range of clinical and lifestyle factors, such as age, BMI, eating habits, physical activity, and hereditary susceptibility. The implementation of thorough preprocessing, feature improvement, and imbalance management sustains the learning process and reduces biased results. The goal of LIME is to provide instance-level explanations that highlight the characteristics that have the biggest impact on each person's risk score. This clarifies predictions. In terms of accuracy, precision, recall, F1-score, and confusion matrices, models generally outperform individual models. Strong generalization across a range of behavioral and demographic characteristics is shown by the ensemble. It ensures that doctors can rely on the results by striking a balance between forecasting accuracy and understandability. The method combines transparency and dependability to allow for efficient, real-time obesity risk assessment. To sum up, it is a reliable instrument for making decisions in medical environments that prioritize prevention.